GPU Over heating over 95C - stable defusion

Hi, I am testing stable defusion that now can run on xavier agx ( 32gb ) to see what it can do and I quickly found out just 5 minutes on 1 picture AGX keeps shutting down because of overheating GPU from JTOP I monitored it show over 95C with fan run at 100% … i try change a few setting with power and fan around still not help same result.

My last try before opening it up is to attach a CPU fan to the side of agx to increase airflow it does not help keep GPU temp around 87 to 90 over 15 mins but this also slows down the work and ultimately shuts down due to thermal problems.

Is there a better way to handle this ? i try to open up and hack around but maybe there is just a simple solution out there I can implement without going extreme.

@DiamondsAI hmm is this a devkit or other system you are using the Xavier module in? I haven’t heard of the devkits overheating/shutting down in normal ambient conditions, but perhaps the hardware guys can comment if there could be an issue with the board itself.

For now, you could change to a lower power model using nvpmodel tool or the power-mode selector widget in the top right of the desktop UI.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.